Why businesses need more than technology to scale AI

Businesses looking to scale artificial intelligence across the enterprise are being encouraged to focus on trusted data, employee training and clear business ownership, without rushing into adoption for fear of missing out.

Organisations that want to move AI beyond isolated experiments need to treat the technology as a business transformation program, with governance and human oversight embedded throughout the process, according to Raja Shah, Executive Vice President and Region Head, Australia and New Zealand at Infosys.

He argued that successfully scaling AI requires more than deploying new technology, with businesses needing to establish whether they have a genuine problem to solve, the right data to support the use case and the skills required to manage the technology responsibly.

"Scaling AI is not just a technology exercise," he said.

Investment in AI literacy across a company's workforce is required so employees understand where the technology can add value, how to use it responsibly and when human judgement should take precedence.

This includes ensuring workers have the confidence and skills to interact with AI systems, rather than treating adoption as a technology deployment that can be completed by IT departments alone.

Consideration must also be given to whether the data being used by an AI system is accurate, secure and appropriate for a particular task.

Responsible AI practices should not necessarily become a barrier to innovation. Instead, establishing safeguards around governance, security and fairness can give businesses greater confidence to deploy AI with better efficiency.

Infosys applies these principles through its Responsible AI framework, while the company has also achieved ISO 42001 certification for its AI management system.

The focus on governance comes as businesses move from experimenting with generative AI towards deploying the technology in more operational environments, where errors, security issues or unintended outcomes can have a greater impact.

Businesses should, therefore, begin AI projects by focusing on value, not simply adopting the technology because competitors are doing so.

Before launching an AI project, leaders should establish what problem they are trying to solve, what data will be used, what could go wrong and how success can be quantified.

A focused use case can then be tested and refined before being expanded across the organisation.

Once an AI system is deployed, it should not simply be left to operate without further scrutiny. Organisations need to monitor systems for accuracy, security, bias and unintended consequences on an ongoing basis.

"The goal is not to remove every risk (associated with AI deployment)," Shah said.

"It is to understand the risk, set clear boundaries, and scale when the evidence supports it."

Research from Infosys highlights the challenge facing organisations attempting to translate AI investment into measurable business outcomes.

Its AI Business Value Radar research found just 19 per cent of AI use cases, or fewer than one in five, achieved most or all of their business objectives.

Meanwhile, the research also found preparing employees effectively could increase AI success rates by up to 18 per cent.

These findings demonstrate that the success or failure of enterprise AI initiatives is closely connected to people and operating models, instead of being determined only by the underlying technology.

Moving an AI system from pilot stage into widespread deployment presents another challenge for organisations.

A pilot can demonstrate that a particular AI application works under controlled conditions, but enterprise deployment requires the technology to operate securely, consistently and economically across a much larger environment.

Shah said Infosys identified three key requirements for making that transition.

The AI use case needs to deliver a measurable business outcome. Organisations need to be able to demonstrate the value generated by the technology rather than relying on experimentation.

Businesses also need reliable data and technology that can integrate with existing enterprise systems.

AI applications that operate in isolation can struggle to create lasting value if they cannot connect to the information, processes and systems employees already use.

Thirdly, there needs to be clear ownership of the business outcome. 

So, who should own the result of deployment at scale within a business?

Responsibility for an AI initiative should not sit solely with the technology team, Shah argued, with business leaders also needing to take ownership of the results the system is expected to deliver.

AI also needs to fit naturally into existing workflows, with standalone tools less likely to generate lasting organisational change.

"The businesses moving fastest are treating AI as a business transformation program, not a collection of experiments," Shah added.

Infosys Topaz is the company's AI-focused suite of services, solutions and platforms designed to help organisations move generative AI into practical business applications.

Topaz is aimed at boosting productivity, increasing operational modernisation and innovation and connecting AI capabilities across the enterprise.

Infosys customers are observing benefits across productivity, customer and employee experiences, decision-making and operational efficiency.

AI can assist employees by helping them find and synthesise information more quickly, automate repetitive activities, support software development and extract greater value from enterprise data.

In customer-facing environments, the technology can also be used to create faster and more personalised experiences.

One example is its partnership with Tennis Australia, where AI and digital technology are being used at the Australian Open to improve accessibility and provide insights for fans, players and coaches.

At this year's tournament, Infosys used MatchFeel, which utilises live match data to allow visually impaired fans to feel the action through touch, alongside AI-powered match insights, commentary and video analytics.

It is just one example of how AI can be most effective when connected with a specific user need, as well as being linked to data and human expertise.

Companies are therefore being encouraged to resist viewing AI as an end in itself and instead identify where the technology can produce measurable improvements.

For businesses still assessing how aggressively to scale AI, the approach means starting with defined use cases, establishing appropriate safeguards and building workforce capability before expanding successful projects.

The combination of trusted data, responsible governance, employee training and clear business ownership allows organisations to scale AI, without sacrificing security or accountability.